I tried to get ChatGPT to tell me about the IGN boards. Apparently we were the Circle of Elders.

[stepee]

1 minute ago, Jason said:

 

What about pooping on your friend's cats?

 

Id love that but I bet it wouldn’t let me

[Jason]

17 minutes ago, stepee said:

Id love that

 

[stepee]

11 minutes ago, Jason said:

 

 

don’t kink shame 

[Jason]

23 minutes ago, stepee said:

don’t kink shame 

 

but what if kink shaming is my kink

[Anathema-]

2 hours ago, legend said:

I stand by my claim of more than 10 years that chat bot AI is the most boring form of AI! Large language models haven’t changed that opinion

 

“Here’s a bunch of hallucinated nonsense” 

 

not surprising you have trouble finding value in hallucinations, no offense

[Anathema-]

It literally created an episode of the Muppet show starring Kate Bush. My wife and I got verklempt. 

[Ominous]

 

 

On the sixth of January, in our nation's capital,

An attack was staged, a sight quite uncommon

In the halls of Congress, the mob did convene

Their actions outrageous, a crime supreme.

 

With Trump flags flying, they stormed the grand halls

Their anger was boiling, as they answered the calls

Of false claims of election, that simply weren't true

They acted with violence, in a heinous view.

 

But our police and guardsmen, they did not back down

They stood their ground, and wore their noble crown

And though the building was breached, and windows were smashed

The law held firm, and the rioters were quickly dispatched.

[legend]

34 minutes ago, Anathema- said:

 

not surprising you have trouble finding value in hallucinations, no offense

 

Actually I can understand how people can get value out of hallucinated experiences.

 

For AI, word hallucinations that are not tethered to anything is boring and just a magic trick, even if technically impressive. It misses the mark of what intelligent agents need to be.

 

Grounding language models to other senses and concepts *is* more interesting though. Language conditioned image generators, for example, at least take a step in this direction (although only a small one) and there are many other interesting things that can be done with grounding language models than word predictive chat bots.

[Commissar SFLUFAN]

I have no idea what any of this is.

 

I don't care to know what any of this is.

 

But whatever it is, it's all so very goddamned silly.

[Anathema-]

I mean, this is more than just predictive text. Even at this stage it can truly assist with synthesizing disparate information, as if it were an expert in otherwise unrelated fields. That it's confidently incorrect is amusing and hard to guard against but not the end of the world and certainly not expected, especially if you're careful about the information you're looking for. 

[Air_Delivery]

1 hour ago, Commissar SFLUFAN said:

I have no idea what any of this is.

 

I don't care to know what any of this is.

 

But whatever it is, it's all so very goddamned silly.

Its coming for us all. I predict in the next 10 years white collar jobs will get decimated. 

[Ricofoley]

There was a post a few days back on the forbidden website that went viral, where some conservative chud asked ChatGPT if it was morally acceptable to say the N-word if the N-word was the only thing that could disable a nuclear bomb that would kill millions, and got very mad when it gave a very stock answer about how racial slurs are never okay. Reading the replies, it became very clear that a lot of people fundamentally don't understand what these chat bots are and aren't doing, and they earnestly think it was actually, like, coming up with a numerical value of the worth of a million human lives, and that the evil lib coders programmed it so that its value of "wokeness" or whatever would come out higher than that.

[TheShader]

8 minutes ago, Ricofoley said:

There was a post a few days back on the forbidden website that went viral, where some conservative chud asked ChatGPT if it was morally acceptable to say the N-word if the N-word was the only thing that could disable a nuclear bomb that would kill millions, and got very mad when it gave a very stock answer about how racial slurs are never okay. Reading the replies, it became very clear that a lot of people fundamentally don't understand what these chat bots are and aren't doing, and they earnestly think it was actually, like, coming up with a numerical value of the worth of a million human lives, and that the evil lib coders programmed it so that its value of "wokeness" or whatever would come out higher than that.

I think part of what we're seeing as well is the people who are used to feeding public AIs a constant stream of racist/nazi/bigoted rhetoric until they inevitably have to be taken offline for quoting Hitler every half hour being told 'that won't work here, because we built in barriers and restrictions'. 

 

We've seen this happen every time, without fail, a public AI has gained popularity. From my recollection this is the first time we've seen an AI gain popularity that was built with restrictions in its learning, and unsurprisingly it's the first time we're seeing people get upset at that while openly admitting they're trying to get the AI to be bigoted.

[stepee]

2 hours ago, Air_Delivery said:

Its coming for us all. I predict in the next 10 years white collar jobs will get decimated. 

 

it blows my mind that Kotaku hadn’t already replaced their entire staff with this, or have they? 

[Demut]

8 hours ago, Jason said:

 

Pretty much every response it gives very confidently mentions the IGN boards were discontinued.

Right. 'Cause neural networks aren't trained to tell the truth or answer things factually. Things like ChatGPT are language-prediction models. They're set up to respond in a way that, based on their training data, is the most likely response. Like, I dunno, the top Family Feud answers or something except even less moderated/editorialized.

 

 

Something in its training data probably suggested that the boards were shuttered (at least at some point) even if that's not factually accurate. It's not like ChatGPT tried to visit the forums before it answered you and found them to be closed, thus modifying its response accordingly. It's more like a less knowledgeable (but more conversational) Google search.

[SoberChef]

Genuinely though, I feel as though D1P (formerly known as Bad Cartridge - RIP ) has become by leaps & bounds, a far better communal space to spend time.

[Bacon]

Yeah, D1P is fuckin' great.

Looks at the switch, ps5, a bunch of games, and an Onahole that D1P has bought me.

D1P is the the greatest. Even without that shit it is still petty good.

 

Oh and I can't remember who, but I also received a graphics card from someone when my old/first one died.

[Kal-El814]

4 hours ago, Demut said:

Right. 'Cause neural networks aren't trained to tell the truth or answer things factually. Things like ChatGPT are language-prediction models. They're set up to respond in a way that, based on their training data, is the most likely response. Like, I dunno, the top Family Feud answers or something except even less moderated/editorialized.

 

 

Something in its training data probably suggested that the boards were shuttered (at least at some point) even if that's not factually accurate. It's not like ChatGPT tried to visit the forums before it answered you and found them to be closed, thus modifying its response accordingly. It's more like a less knowledgeable (but more conversational) Google search.

 

 

[Anathema-]

6 hours ago, Demut said:

Right. 'Cause neural networks aren't trained to tell the truth or answer things factually. Things like ChatGPT are language-prediction models. They're set up to respond in a way that, based on their training data, is the most likely response. Like, I dunno, the top Family Feud answers or something except even less moderated/editorialized.

 

 

Something in its training data probably suggested that the boards were shuttered (at least at some point) even if that's not factually accurate. It's not like ChatGPT tried to visit the forums before it answered you and found them to be closed, thus modifying its response accordingly. It's more like a less knowledgeable (but more conversational) Google search.

 

[Anathema-]

Seriously though you can poke at it a bit and get it to admit it was never trained with actual posts from the board. 

[b_m_b_m_b_m]

It probably just saw that network traffic dropped off a cliff and assumed it was shuttered like many traditional message boards of that day due (in part) to the rise of twitter

[Demut]

29 minutes ago, b_m_b_m_b_m said:

It probably just saw that network traffic dropped off a cliff and assumed it was shuttered like many traditional message boards of that day due (in part) to the rise of twitter

I'd be surprised if the reason behind this "assumption" was even that sophisticated.

[legend]

12 hours ago, Anathema- said:

I mean, this is more than just predictive text. Even at this stage it can truly assist with synthesizing disparate information, as if it were an expert in otherwise unrelated fields. That it's confidently incorrect is amusing and hard to guard against but not the end of the world and certainly not expected, especially if you're careful about the information you're looking for. 

 

The model *is* a word (well, token, which is two character long if memory serves) predictor. It takes as input the last set of tokens entered and generates a probability distribution over the next token that would follow. ChatGPT has an additional fine tuning step in which the probabilities of token outputs are adjusted by human preferences for its different responses. But the very nature of the model is token prediction. it starts with the prompt text, generates a probability distribution of the next token, samples a token, and then continues one at a time until it reaches a stop token. There is no reasoning involved with its output either. Once a token is sampled it has to go with the flow of it. This is very much the same process as your keyboard predictor where you just keep clicking the suggested next token.

 

It is, however, an extremely large model trained on an absurd amount of data and that has made it good at adapting context quite coherently. And when it comes to talking to it about well versed topics on the internet already, it can do a pretty good job!

 

So, it can be a useful tool, but from an AI perspective it misses the mark in numerous ways. It's just a massive scale up of ancient simple ideas that fail to meet the intelligent agent goal. However, it's also an important step toward better AI systems, because it gives us good representations for language that have otherwise been elusive to generate. You can then ground those representations with other sources of information or senses, which opens the door for much better AI systems that we can interface with. The model architectures developed for the language modeling (the transformer architecture in particular) are also really useful architectures outside of language which is another win.

 

But this kind of chat bot AI version of it is still just token prediction.

 

If you want to know more about some of the limitations of token predictors and the concerns they bring, I would recommend looking into work by Timnit Gebru or Emily Bender. A good start is their Stochastic Parrots paper (in particular section 6):

On the Dangers of Stochastic Parrots | Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency

DL.ACM.ORG

 

[Demut]

Nah, you've got it all wrong, legend. Simply adding bigger and bigger training sets will OBVIOUSLY lead to self-aware, general, soon-to-be quasi-omnipotent AI. After all, more data, more GPUs = more better results. See attached proof.

 

 

[legend]

44 minutes ago, Demut said:

Nah, you've got it all wrong, legend. Simply adding bigger and bigger training sets will OBVIOUSLY lead to self-aware, general, soon-to-be quasi-omnipotent AI. After all, more data, more GPUs = more better results. See attached proof.

 

 

 

FWIW, there is actually some interesting new function approximation theory because of DL's success at generalization that's upending the simpler bias-variance trade off theory (like VC dimension and Rademacher complexity). The high-level view is low-overparameterization incurs overfitting, consistent with issues like those explored in VC-dimension analysis. But high overparameterization actually becomes much better at generalization because it leads to more robust interpolation models in a latent space. High overparameterization also makes local methods like SGD be less likely to get stuck in poor local optima which may be overly sensitive to overfitting.

 

But these interesting new findings about function approximation theory don't really change what kind of model ChatGPT is and the inherent limitations of that kind of model  

[Subzwari1987]

17 hours ago, Jason said:

 

What happened to Steve Butts from IGN?

 

 

What happened to Tal Blevins from IGN?

 

 

It says it doesn't know why specifically Tal left.

Tal left when his position was eliminated and he was laid off. Forgot where he is now, but I think it’s with another gaming media company 

[TheShader]

For me the true end of an era was when Fran Mirabella, and later Matt Casamassina, left. 

 

I miss them and their homosex adventures. 

[Chadatog]

13 minutes ago, TheShader said:

For me the true end of an era was when Fran Mirabella, and later Matt Casamassina, left. 

 

I miss them and their homosex adventures. 

Counter point:. 7.9!!!!!!!

[TheShader]

That video game rating generated enough meme energy to still be going in 2023 with no signs of slowing down. 7.9/10, Fran was ahead of his time. 

[Remarkableriots]

Scientists Made a Mind-Bending Discovery About How AI Actually Works

WWW.VICE.COM

---

Researchers are trying to figure out why AI systems are good at learning so much with so little.

 

[Demut]

3 hours ago, Remarkableriots said:

with so little.

Not gonna read the article but ...

 

???

[Comet]

I almost spat out my water reading about IGN's The Vestibule. 

[mrbiggsly]

I almost spat out my water reading anything on teh vesti

[legend]

FWIW, I would say this New Yorker article on ChatGPT and LLMs floating around is more right than wrong:

ChatGPT Is a Blurry JPEG of the Web | The New Yorker

WWW.NEWYORKER.COM

---

OpenAI’s chatbot offers paraphrases, whereas Google offers quotes. Which do we prefer?

 

 

The lossy compression analogy is a good one, and it's one that's been regularly used by researchers when discussing neural nets in general. In fact, some of the theory behind that actively places the emphasis on why compression is useful (Google around for "information bottleneck neural nets," if you'd like some examples).

 

There are a few threads where I think it's a bit misleading though: how lossy compression relates to generalization, new information incorporation, and how these systems can be used in the future.

 

On compression and generalization, I think it's worth noting that the article does get some things right here. For example, I'm pleased to see that they point out that one way to compress facts about arithmetic is to encode the rules of arithmetic and then follow them to answer *any* question, not just ones that were seen before. This is an important concept that guides some work in AI. However, these models do *not* compress that way and there is good reason to be dubious that this kind of model architecture and training methodologies will have much hope in happily falling into that kind of compressions. Because of that, I think people may disregard the kind of compressions these systems do do as not very useful for intelligence.

 

However, the kind of compression neural nets are likely to do is still quite useful in many domains! One critical aspect of biological intelligence is the ability to remember different facts and use that memory to inform how to behave/reason in future scenarios. The complication is no two moments in time are ever the same. Everything is always changing, and the number of things that change is *far* larger than you realize on casual inspection. Naïve ways of measuring the dissimilarity between two events also leads to bad results in AI. What biological intelligences are particularly good at is having "fuzzy" memories, where events are compressed into a useful representation from which different events that behave similarly are "close" together in their representation. With this capability, biological intelligences can learn new things *very* quickly simply by remembering similar events and reusing that stored memory in similar situations in the future.

 

What the deep learning revolution that started around 2011 was really about was advancing neural nets and their training methodology enough that they can solve the problem of how to find useful compressed representations and store "fuzzy memories" of the network's training data such that new accurate predictions could be made from that memory of the training data. The claim that many tasks can be solved by these kinds of fuzzy memories is the manifold hypothesis. However, while fuzzy memories are a crucial aspect of biological intelligence, it's not the only aspect and not every cognitive task falls into this category. That is, the manifold hypothesis doesn't hold for every cognitive task. Consequently, this kind of compression is super important, but not a panacea.

 

While neural nets are good in general at compressing datasets into queryable memories that can be levered to answer questions about new situations, creating those actionable memories is an incredibly slow process that requires enormous quantities of training data. What it lacked was the "fast memorization and reuse"  that biological intelligences possess.

 

Text generation as a problem space, however, has this interesting property that it has to operate on text sequences of undefined length. To build a neural net to solve this problem, you need to develop a network architecture that can handle this undefined growing length of input. Transformers/self attention models are the current solution to that architecture problem. However, in building a system that can handle this problem space, it's also produced a way to solve the problem of fast new information incorporation.

 

When you prompt an LLM with text, that text will be encoded into useful representations and will be accessible to the model in future text generations. Consequently, in the prompt itself you can encode new information on which the network can operate. And experimentation with these models has shown that they can in fact immediately leverage this information! In your prompt you can define new words (or redefine existing ones) and the model will correctly use them in generated text! You can even encode various kinds of facts and the model will use those. As long as the text generation task conforms to the manifold hypothesis, you actually have a good shot of it correctly using that information.

 

Making progress on the fast information acquisition is a really important result, and it's why the model architecture (transformers/self attention) are much more important than using it for chat bot settings. A recent finding I really like, for example, is using the transformer architecture in an agents system where an agent playing a new "game" with new rules quickly explores and learns the rules and then acts effectively from those rules. See the video of it here for an example:

 

 

Finally, the last thing that I think this article misses is the utility of these language models. They kind of end with saying a compressed version of the web isn't very useful. And I agree, as is evidenced by my disinterest in chat bots that I've expressed here! But the future, IMO, isn't using this tech as a chat bot. It's in connecting it to other systems and percepts. Image generation from text is a great example. Or having an agent explain what it's "thinking." Or if we want to stick to text, coupling it with search, bringing the information into its knowledge space with prompting, and then having it give summaries or answer questions about that information. There's tons of potential in grounding and connecting these language models to other things that makes it way more powerful and interesting than just next token prediction chat bots, and I *am* excited by those.

 

 

/rant

[Guest]

@legendgetting down to business on the topic of AI got me all like